Thinking Machines Corporation
Thinking Machines Corporation was an American supercomputer and artificial-intelligence company that commercialized massively parallel computing through its Connection Machine systems.
Last updated October 14, 2026
AI crawler reads
No AI crawler has read this profile since Oct 28, 2026.
Overview
Thinking Machines Corporation was an American high-performance-computing company founded in 1983 by Sheryl Handler and W. Daniel "Danny" Hillis. The company was created to commercialize Hillis's Massachusetts Institute of Technology research on massively parallel computer architectures. Its principal product family, the Connection Machine, used large numbers of processors operating in parallel and became one of the most recognizable supercomputing platforms of the late 1980s and early 1990s. The company began in Waltham, Massachusetts, and moved in 1984 to Kendall Square in Cambridge, near MIT's artificial-intelligence research community. Its early systems, including the CM-1 and CM-2, used thousands of relatively simple bit-serial processors arranged through a hypercube interconnect. These machines were designed for workloads that could be divided into many simultaneous operations, including scientific computing, artificial intelligence, image processing, simulation, and data analysis. Programming environments included *Lisp, CM Lisp, C*, and CM Fortran, supported by proprietary compilers and a specialized parallel instruction set. Thinking Machines expanded the product family with the CM-200, CM-5, and CM-5E. The later CM-5 generation moved from the earlier single-instruction, multiple-data architecture toward a multiple-instruction, multiple-data design that combined commodity SPARC processors, proprietary vector processors, and a high-bandwidth fat-tree network. The company also introduced DataVault, an early commercial RAID 2 disk-array product, and developed parallel software and data-mining tools for its own systems and other high-performance computers. The company became profitable in 1989, helped in part by contracts from the Defense Advanced Research Projects Agency. In 1990 it reported approximately $65 million in hardware and software sales and was described as the market leader in parallel supercomputers. Connection Machines achieved considerable visibility in government, research, and popular culture, while the company competed primarily with Cray Research and also faced challenges from nCUBE, Kendall Square Research, MasPar, Meiko Scientific, and other parallel-computing vendors. By 1993, four of the world's fastest computers were Connection Machines, according to contemporary descriptions of the industry. Its growth depended significantly on government and research procurement. In 1991, DARPA and the U.S. Department of Energy reduced purchases amid criticism that government contracting practices favored Thinking Machines over competing suppliers. Export restrictions also limited the international sale of the company's most powerful systems. The resulting market pressure, combined with broader financial difficulties, contributed to losses by 1992. Sheryl Handler was forced out as chief executive during this period. Thinking Machines filed for Chapter 11 bankruptcy in August 1994. Sun Microsystems acquired the company's hardware business, while the reorganized company continued for a time as a smaller software business focused on parallel-programming tools, cluster software, and data-mining products for installed systems and former competitors' machines. Sun acquired the parallel-software development operation in December 1996. Thinking Machines subsequently continued as a data-mining company until Oracle acquired it in 1999. Oracle later acquired Sun Microsystems, bringing together substantial portions of the company's technological legacy under a common corporate owner. The company's influence extended beyond its direct products. Its staff and alumni helped seed later parallel-computing and data-integration ventures, and several former employees joined Sun's development of parallel enterprise systems. Thinking Machines also hosted development of the Wide Area Information Server, or WAIS, associated with Brewster Kahle; that work later influenced the Internet Archive and related digital-preservation projects. Although the company no longer operates as an independent business, the Connection Machine remains an important milestone in the commercialization of massively parallel computing.
History
Thinking Machines Corporation was established in 1983 in Waltham, Massachusetts, by Sheryl Handler and W. Daniel "Danny" Hillis. Hillis had developed massively parallel computing concepts during his doctoral work at MIT, and the company was formed to transform those ideas into commercially usable systems. In 1984 it moved to Cambridge's Kendall Square, placing the business near MIT and the region's artificial-intelligence and computer-science research community. The company's central product was the Connection Machine. The first generations, the CM-1 and CM-2, employed 65,536 small bit-serial processors, with smaller configurations also offered. Their processors were connected through a hypercube network intended to reduce communication distance and latency. The architecture was particularly suited to problems that could be decomposed into many concurrent operations. The CM-1 through CM-200 followed a single-instruction, multiple-data model, while the CM-5 and CM-5E adopted a multiple-instruction, multiple-data design using SPARC processors, proprietary vector processors, and a fat-tree interconnect. Programming the systems required specialized software. Thinking Machines supported *Lisp and CM Lisp for symbolic and artificial-intelligence applications, C* for parallel C-style programming, and CM Fortran for scientific workloads. Proprietary compilers translated these languages into instructions for the Connection Machine architecture. The company also developed DataVault, an early commercial RAID 2 disk array, and later produced parallel software and data-mining tools that could be used with commodity clusters and other parallel computers. Thinking Machines grew rapidly during the late 1980s. It became profitable in 1989, with DARPA contracts contributing to its business, and generated approximately $65 million in hardware and software sales in 1990. It was regarded as a leading parallel-supercomputer supplier, competing with Cray Research as well as nCUBE, Kendall Square Research, MasPar, and Meiko Scientific. By 1993, four of the world's fastest computers were reported to be Connection Machines. The systems were used in government, research, artificial intelligence, scientific analysis, and other computationally intensive settings. The company's dependence on government and research procurement became a vulnerability. In 1991, DARPA and the Department of Energy reduced purchases following criticism that federal support and procurement decisions favored Thinking Machines over rivals. Export controls also limited the overseas distribution of the most capable systems. By 1992, Thinking Machines was losing money, and Handler was removed from the chief executive position. The company filed for Chapter 11 bankruptcy in August 1994. After the filing, Sun Microsystems acquired the hardware portion of the business. Thinking Machines re-emerged in reduced form as a software company, concentrating on parallel-programming tools, cluster-oriented software, and data-mining products for its installed base and users of competing parallel machines. Sun acquired the parallel software development division in December 1996. The company remained active as a pure data-mining business until Oracle acquired it in 1999. Oracle's later acquisition of Sun Microsystems reunited significant portions of Thinking Machines' former technology and personnel lineage. The company's legacy was distributed through its products, intellectual property, and employees. Former Thinking Machines staff helped establish or develop later parallel-computing and data-integration companies, while others joined Sun and contributed to parallel enterprise-computing products. Brewster Kahle's work on the Wide Area Information Server at Thinking Machines also became part of the technical history leading to the Internet Archive and related preservation initiatives. The company's machines appeared in popular culture, including references and visual appearances in films and novels, reinforcing its identity as a symbol of advanced computing even after its commercial collapse.
- 1999
Acquired by Oracle
[1]Oracle acquired the remaining data-mining company.
- 1996
Parallel software operation acquired by Sun
[1]Sun Microsystems acquired the company's parallel software development section in December.
- 1994
Files for Chapter 11
[2]Thinking Machines filed for bankruptcy protection in August, after financial losses and market pressures.
- 1993
Connection Machines occupy leading supercomputer positions
[1]Four of the world's fastest computers were reported to be Connection Machines.
- 1990
Reaches parallel-computing market leadership
[1]The company sold approximately $65 million in hardware and software and was described as the parallel-supercomputer market leader.
- 1989
Becomes profitable
[1]The company reached profitability, supported partly by DARPA contracts.
- 1985
Registers think.com
[1]Thinking Machines registered one of the earliest .com domains, think.com.
- 1984
Moves to Kendall Square
[1]The company moved from Waltham to Cambridge, Massachusetts, near MIT's technology and artificial-intelligence community.
- 1983
Company founded
[1]Sheryl Handler and W. Daniel Hillis founded Thinking Machines Corporation to commercialize massively parallel computing research.
Products and positioning
A pioneering high-performance-computing company focused on massively parallel architectures, scientific and artificial-intelligence workloads, and specialized software for exploiting large processor arrays.
Massively parallel supercomputer
CM-1
The first Connection Machine model commercialized Thinking Machines' massively parallel architecture. It used a large array of bit-serial processors connected through a hypercube network and was intended for highly parallel scientific, artificial-intelligence, and data-processing workloads.
Massively parallel supercomputer
CM-2
The CM-2 extended the early Connection Machine design and was offered in configurations based on 65,536 processors, with smaller versions also available. It used the company's SIMD architecture and specialized software environment for parallel computation.
Massively parallel supercomputer
CM-200
The CM-200 was a later SIMD member of the Connection Machine family. Like its predecessors, it relied on a highly connected processor array and proprietary compilers and languages to execute workloads divided across many processors.
Massively parallel supercomputer
CM-5
The CM-5 represented a major architectural change for the company. It used a MIMD design combining commodity SPARC processors, proprietary vector processors, and a fat-tree network, making it suitable for a broader range of high-performance applications than the earlier SIMD systems.
Massively parallel supercomputer
CM-5E
The CM-5E was a later Connection Machine configuration based on the CM-5 approach to distributed parallel processing. It formed part of the company's effort to provide scalable systems using a mixture of commercial processors and specialized interconnect technology.
RAID disk array · 1988
DataVault
DataVault was an early commercial RAID 2 disk-array product introduced by Thinking Machines around the late 1980s. It addressed storage reliability and throughput needs associated with high-performance computing installations.
Parallel-computing software
Parallel software tools
After the bankruptcy reorganization, Thinking Machines continued developing software for parallel programming and commodity clusters. These tools supported users of the company's installed base as well as users of other parallel-computing systems.
Data-mining software
Darwin
Darwin was a data-mining toolkit associated with Thinking Machines' Business Supercomputer Group. The toolkit was later acquired by Oracle, while much of its development team had moved to other companies after the bankruptcy.
Information-retrieval software
Wide Area Information Server
The Wide Area Information Server, or WAIS, was developed at Thinking Machines by Brewster Kahle. It became an influential early information-retrieval technology and formed part of the technical background to later Internet Archive-related work.
Brand decisions
1994
Chapter 11 restructuringOtherBackground
The company was losing money after government purchases declined, export restrictions constrained international sales, and competition in parallel computing intensified.
Action taken
Thinking Machines filed for Chapter 11 bankruptcy protection in August 1994 and separated its hardware activities from a smaller continuing software business.
Aftermath
Sun Microsystems acquired the hardware business, while the reorganized software operations continued until Sun acquired the parallel software development section in 1996 and Oracle acquired the remaining data-mining business in 1999.
Sources2 sources
Cite this profile: Cite the canonical profile. /brand-wiki/thinking-machines-corporation · Editorial policy · How profiles are compiled